Data Sea; Data Do

We live and work in a sea of data trying to envision, understand, and plan for the future. That data, however, by its very nature is primarily about the past.


In order to make use of data in ways applicable to the future, we apply theories. These theories are often referred to as predictive models, which are frequently integrated into decision models along with other theoretical constructs such as descriptive models. The process of applying theories or models with the future in mind is predictive analysis.


This kind of analysis is employed daily across a diverse collection of organizations relying on disparate and common data sources from both internal and external sources. Its most widespread application lies in organizations such as actuarial science, financial services, insurance, telecommunications, retail, travel, healthcare, pharmaceuticals, and other fields including portfolio and program management. 


Using data coupled with theories or models to predict the future is more than simply "common," it is ubiquitous and the foundation for strategic, tactical, and operational processes in our governments, commercial enterprises, and other organizations, as well as in our personal lives on a daily basis. We humans are constant predictors but our results are less than consistently predictable. 


Much attention is focused today on data: Big data; a sea of data. Obtaining more and more data is becoming easier and easier and it's coming at us faster and faster. Yet predicting the future is not appreciably more accurate today than ten years ago. Simply acquiring more--even more accurate--data isn't necessarily translating into better results. Why? 


Even the most accurate data (historical, current, or real-time), even in overwhelmingly huge amounts, when viewed through the lens of a flawed theory or model, becomes distorted, twisted and useless. In other words, perfect data produces imperfect results when analyzed using even slightly imperfect theoretical constructs.


At least as much--or more--attention is needed on envisioning, creating, testing, and implementing valid analysis techniques and models as is given to data acquisition. Without advances in models, constructs, and analysis techniques, acquiring more data leads to erroneously high levels of confidence but no greater accuracy in predictive results.

Posted by: William W. (Woody) Williams

Subscribe